new drug development d roses , 2, ann m saunders , michael ...drsmith/migrated/metx270/html/roses et...

New applications of disease genetics and pharmacogenetics todrug development§

Allen D Roses1,2, Ann M Saunders2, Michael W Lutz2, Nanyin Zhang3,Ahmad R Hariri2, Karen E Asin4, Donna G Crenshaw1, Kumar Budur4,Daniel K Burns1 and Stephen K Brannan4

Available online at www.sciencedirect.com

ScienceDirect

TOMMORROW is a Phase III delay of onset clinical trial to

determine whether low doses of pioglitazone, a molecule that

induces mitochondrial doubling, delays the onset of MCI-AD in

normal subjects treated with low dose compared to placebo.

BOLD imaging studies in rodents and man were used to find the

dose that increases oxygen consumption at central regions of

the brain in higher proportion than activation of large corticol

regions. The trial is made practical by the use of a

pharmacogenetic algorithm based on TOMM40 and APOE

genotypes and age to identify normal subjects at high risk of

MCI-AD between the ages of 65–83 years within a five year

follow-up period.

Addresses1 Zinfandel Pharmaceuticals, Inc., Durham, NC, United States2 Duke University Medical Center, Department of Neurology, Durham,

NC, United States3 Pennsylvania State University, Department of Biomedical Engineering,

State College, PA, United States4 Takeda Pharmaceuticals, Deerfield, IL, United States

Corresponding author: Roses, Allen D ([email protected],

[email protected])

Current Opinion in Pharmacology 2014, 14:81–89

This review comes from a themed issue on Neurosciences

Edited by David G Trist and Alan Bye

For a complete overview see the Issue and the Editorial

Available online 3rd January 2013

1471-4892/$ – see front matter, # 2013 The Authors. Published by

Elsevier Ltd. All rights reserved.

http://dx.doi.org/10.1016/j.coph.2013.12.002

It is important when discussing genetics and pharmaco-

genetics [PGX] in clinical pharmacology and drug de-

velopment to acknowledge the technologies that are

commonly practiced in drug development. From a

genetics point of view, the contents of clinical pharma-

cology databases for drug metabolites and PK/PD have

certainly been extended by genetic/genomic technol-

ogies over the past two decades. In particular, sequencing

§ This is an open-access article distributed under the terms of the

Creative Commons Attribution-NonCommercial-No Derivative Works

License, which permits non-commercial use, distribution, and repro-

duction in any medium, provided the original author and source are

credited.

www.sciencedirect.com

of highly polymorphic loci such as the HLA region and

genes involved in drug metabolism have increased speci-

ficity of genetic associations with pharmacokinetic or

pharmacodynamic phenotypes. This expanded compen-

dium of polymorphisms, including less frequent poly-

morphisms, has increased our ability to explore the

genetics of uncommon events in different ethnic groups.

One example is the genetic underpinnings of the hyper-

sensitivity syndrome observed in about 4% of Caucasian

users of abacavir to treat HIV infection [1]. GlaxoWell-

come received an accelerated approval to market abaca-

vir, but approval was accompanied by an expectation by

the FDA and EMA that the company would investigate

the hypersensitivity syndrome. Cooperation throughout

clinical development at GlaxoWellcome was necessary to

undertake the experiments that would allow for the

identification of the subgroup of patients for whom aba-

cavir was contraindicated. While the HLA-B57 locus was

known in 1997, from a database point of view it took

several years to identify the B-5701, 5702 and 5703

polymorphisms at the locus using genomic sequencing.

The risk of the hypersensitivity reaction was present only

in patients with the B-5701 allele.

Today human drug trials provide opportunities to identify

efficacy responders and to characterize their genetics

compared to non-responders [2]. Since very large num-

bers of patients are usually not available [affordable or

practical] as they may be for studies of risk factors for

common diseases studies, the analytical process is a bit

reversed from searching for genetic mutations for Men-

delian disease diagnoses. Patients must be identified and

followed during and after treatment. Predicting efficacy

for individual patients who might also be at risk for an

adverse response will take time to become common in

medicine. Drug trials take years and must be designed to

prospectively evaluate patient responses. A basic require-

ment for genetic testing is that consented DNA samples

must be collected across the entire study. In reality, to

accomplish collection of DNA from all study participants

requires planning and patient consent, without which

pharmacogenetics cannot be used effectively for hypoth-

esis generation, hypothesis testing and for regulatory

decisions.

It is a bit different for uncommon adverse events where

the genetic loci associated with relatively rare, overt


[email protected]

[email protected]

http://www.sciencedirect.com/science/journal/14714892/14



http://www.sciencedirect.com/science/journal/14714892

82 Neurosciences

phenotypes are typically easier to identify. For safety

PGX, the uncommon mutations can be assessed against

population samples, but for efficacy within a trial it is

necessary to study all participants so that efficacy and lack

of efficacy can be compared in participants who received

drug. It is, however, a problem when DNA is not collected

or unavailable in individuals with adverse events. Bank-

ing DNA samples from important studies is still quite

uncommon in the industry [1,14].

The goal of efficacy PGX is to find a genetic biomarker

that predicts high expectation of efficacy in a sufficient

number of people to qualify the test for selection of drug

therapy. These studies focus on the response to treat-

ment, and may be quite distinct from genetic diagnostic

tests for disease. Beginning with any early Phase II

treatment trial, the objective is to differentiate those

treated individuals who respond to the drug but are not

responsive to placebo. This can initially be evaluated in

small studies during which defined clinical responses are

measured. The more obvious the efficacy response, the

more certain early associations can be made. Efficacy

PGX depends on the choice of therapeutic agent, dose,

and the responses of the individuals receiving the drug. It

is also influenced by the type of genetic search performed.

Candidate gene lists involving proposed disease pathways

have provided an effective starting point for translation to

clinical trials for several decades. Genome-wide associ-

ation studies (GWAS) have produced many gene lists

with subsequent rationales provided for selected genes.

To date, there are few examples where specific genes

‘‘discovered’’ by GWAS lists have translated into clini-

cally relevant programs, outside of oncology where

affected tissue is more available to confirm with gene

expression studies.

Each subsequent Phase II trial will provide greater stat-

istical support for markers associated with drug efficacy.

Recently, an investigation of highly polymorphic struc-

tural genes that carry highly polymorphic variations at a

single locus compared to SNPs has reduced the time

necessary for translation to clinical trials [3]. A high

degree of polymorphic variations contribute to the pro-

portion of individuals in a population who are informative

at that locus. Efficacy biomarkers identified and validated

in Phase II may be useful in proof of concept studies and

to stratify and improve the efficiency of Phase III trials.

PGX markers that enrich for responders in Phase III

studies may allow for a more robust efficacy signal and

a faster path to registration. Companion diagnostics can

be also be qualified in late clinical trials to identify

individuals with a higher ‘risk’ of a beneficial response

[3,4]. It should also be noted that polymorphisms in one

ethnic group may be absent or occur at a lower allele

frequency in other ethnic groups. This emphasizes the

need to examine genetic markers carefully in different

ethnicities, especially with highly variable markers [5�].


This short opinion piece will emphasize some of the

newer genetic technologies that can facilitate Phase III

trial design. Efficacy PGX is a relatively new application

of genetics in clinical development because it requires

DNA collection during the trial. In this new paradigm for

drug development, drug discovery becomes more de-

pendent on studies that utilize these PGX associations.

Safety PGX in the aforementioned abacavir trial demon-

strates the translation of a genetic marker to ensure safer

use of the medication. In addition, economics are often

the major consideration for use of a test, particularly for

defining efficacy for reimbursement. Predicting adverse

responses also leads to extended market use in addition to

enhanced safety long after patent exclusivity for the drug

compound expires. In the case of abacavir it is still used in

HIV drug combinations for HLA-B5701 negative

patients, thereby extending its commercial value after

its chemical patents expired.

New technologies and innovation in a Phase IIIclinical trial: trial design to study delay ofonsetCentral nervous system [CNS] diseases have recently

experienced declining interest by major drug developers

because the costs of doing clinical studies in these disease

areas can be high and likelihood of approval low. The

basic premise that early clinical pharmacology and safety

studies can rule out [kill fast] compounds works for Phase

I, but there are major differences in the cost of Phase II

and Phase III clinical studies. It can be very costly to run

clinical trials to observe efficacy in a dose-finding trial.

This is especially true for longer trials with clinical end-

points followed over months, such as those performed for

neuropsychiatric disorders.

Studying clinical pharmacology in animals and translating

these findings into human studies is critical for generating

information on dose and trial design that can translate into

reducing overall drug development cost and time to

approval. Imaging technologies have now been applied

to measure effects of different drugs and drug doses on

the pharmacodynamic response of specific anatomical

areas in the brain. While drug dosages are usually pushed

to the highest tolerated level to ensure that efficacy will

not be missed, the chances for adverse events are raised as

well. For neurological diseases, imaging studies in

animals provide a way to determine the most reasonable,

effective dose without waiting for expensive clinical ef-

ficacy studies to define it.

PET and fMRI imaging studies in animals and man can

now be used for dose finding studies in late discovery and

early development, particularly in the neuropsychiatric

diseases from which major pharmaceutical companies

have generally withdrawn over the past five years. These

data can quickly and economically define the lowest dose


New applications of disease genetics and pharmacogenetics Roses et al. 83

with activity in specific regions of the brain, allowing

dose-finding studies to suggest lower, and generally safer,

brain-active doses in a shorter period of time. This is

particularly relevant to disease prevention studies, where

a drug is used to treat unaffected people at high risk of

developing a particular disease. Dose selection for the

drug that is being used in a delay of onset of Mild

Cognitive Impairment due to Alzheimer’s Disease

[MCI-AD] trial, the TOMMORROW study, was

informed by a BOLD MRI study.

The genetic, imaging and phylogeneticmapping basis for translation to theTOMMORROW Phase III clinical trialThe TOMMORROW clinical trial, a large, Phase III trial

to assess delay of onset of MCI-AD in cognitively normal

subjects between 65 and 83 years of age at entry, was

designed to qualify a predictive genetic biomarker in the

context of delaying the onset of MCI-AD using a drug

that may work by increasing the number of mitochondria

in cells. Animal BOLD imaging, which measures anato-

mically specific oxygen utilization during cerebrovascular

response, was used to identify the lowest dose of piogli-

tazone that would activate metabolism in the central areas

of the brain that are associated in humans with clinical AD

[34��] (Figure 1).

In order to translate these animal model data to humans,

functional BOLD imaging was conducted following a

memory test in people who received 14 days of pioglita-

zone treatment. BOLD fMRI responses were used for

dose finding studies of pioglitazone using a within-subject

BOLD fMRI paradigm targeting episodic memory-

related hippocampal activation (Figure 2).

Figure 1

Baseline

0 0.5

Post-dosing

Control D

An illustration of the experimental results reported in a dose finding experim

connectivity between the CA1 region of the hippocampus and the ventral tha

of pioglitazone, far below the dose usually recommended to treat diabetes m

well below human doses for diabetes mellitus, has diminished central activi


These BOLD data led to the choice of drug dosage of

0.8 mg/kg for the Phase III TOMMORROW study. The

TOMMORROW trial is designed to qualify the genetic

biomarker, an algorithm composed of TOMM40 and

APOE genotypes and age at study entry, used to identify

those at high risk to develop MCI-AD and simultaneously

test the efficacy of pioglitazone to delay the onset of this

condition in high risk subjects. (Delay of onset trials are

frequently referred to as prevention trials, but to establish

prevention of the condition would require a much longer

study than the 5-year treatment period that is anticip-

ated.) This daily dosage is less than 3% of the dosage

approved for the treatment of type 2 diabetes mellitus.

Pioglitazone boasts a well-documented tolerability record

of more than 22 million man-years of clinical use at the

doses marketed for treatment of type 2 diabetes mellitus

(15–45 mg daily). A lower dose would seem to be safer for

use in normal individuals whose TOMM40 and APOE

genotypes, and age would suggest a higher risk of MCI in

the next five years.

Other genetic technologiesIt is important to delineate the other genetic technologies

that led to the TOMM40, APOE, and age-at-entry algor-

ithm that is being qualified. Several years ago, we

suggested using molecular evolutionary analysis [phylo-

genetic mapping] to describe the complex genetics of

neurodegenerative diseases that have a component of

inherited risk to identify clinically useful biomarkers

and disease-associated variants. Across the many GWAS

studies for complex diseases, there were floods of newly

identified ‘disease genes’ with small to moderate (odds

ratio of 1.1–2.0) effect sizes. Still, several years later, many

of the variants identified by GWAS have not been

= 0.04 mg/kg D = 0.32 mg/kg

Current Opinion in Pharmacology

ent in awake rats using resting state BOLD MRI. Increased functional

lamus and hypothalamus is seen following 7 daily doses with 0.04 mg/kg

ellitus [30–45 mg daily]. Note that the dose of 0.32 mg/kg, which is still

ty.


84 Neurosciences

Figure 2

Placebo 0.6 mg 6.0 mg

12

10

8

6

4

2

0


An illustration of the results of the human fMRI experiment, demonstrating the activation of oxygen utilization in central regions of the brain usually

affected later by AD. Using this within-subject pharmacologic BOLD fMRI paradigm while testing episodic memory hippocampal activation in the left

hippocampal area was observed, again at a very low dose. At higher pioglitazone doses there is more cortical activity present and less hippocampal

activation observed (Zeineh et al., Science 2003).

translated into actionable findings for disease risk predic-

tion or PGX [6,7��,8��].

Association of APOE4 with increased risk and earlier age

of onset of late-onset AD was first reported in 1993

[9�,10��]. At this time, the association was widely criti-

cized by AD genetic experts in national and international

meetings, sometimes with presented data later found to

be technically in error. Published reports of relatively

small clinical studies confirmed the association between

APOE genotype and risk of AD [11–16]. Since this time,

multiple GWAS have confirmed that the region contain-

ing APOE is highly significantly associated with risk of

developing AD; spectacular p values were reported for the

association of TOMM40 SNPs within the LD region

containing the APOE gene, for example, in the study

of Harold et al. [17] with four TOMM40 SNPs [including

rs2075650, P = 1.8 � 10�157].

Because the associated SNPs in TOMM40 were located

within the LD region containing APOE, TOMM40’s

strong association with AD was attributed to APOE; that

is, the TOMM40 SNPs were correlated with either the

e4, e3 or e2 epsilon allele SNPs of APOE. In reality, if the

association of APOE4 to AD had not preceded these

GWAS data by more than two decades, TOMM40 may

have been identified as the AD gene instead of APOE

[18,19].

Contrast this with the genetics of the amyloid precursor

protein [APP] region on chromosome 21. Even though

there are rare, autosomal dominant APP mutations, there

is no GWAS support for the amyloid precursor protein

[APP] region on chromosome 21 for sporadic, late-onset

AD [20]. These rare mutations in APP that result in early

onset, autosomal dominant AD have been used, in part, to

justify the huge investment in amyloid research and


amyloid-associated mechanisms for drug development

programs for late-onset AD [LOAD]. A pathogenetic

mechanism involving these specific APP mutations has

yet to be defined. However, independent of any APP

mutation, it was observed in 2005 that APP protein

truncated at the carboxyl terminus accumulates in the

Tom40 mitochondrial import channel in the brains of AD

patients and this interaction results in mitochondrial dys-

function [21��,22�].

Early positron emission tomography [PET] data from

several centers demonstrated decreased glucose metab-

olism in brain regions associated with AD pathology in

normal individuals carrying one or both APOE4 alleles

compared to non-APOE4 carriers[23��,24�,25]. Since the

brain depends on glucose, oxygen and the blood circula-

tion needed to transport them, we investigated a hypoth-

esis that there was something wrong with glucose energy

efficiency and in 1996 began looking at drugs that

increased the number of intracellular mitochondria as

potential treatments [26��].

In 1998, SNPs from the PEREC-1 gene next to APOE

within the LD region were determined to be associated

with risk for AD [27�]. At that time a function for the

PEREC-1 gene was unknown. In 2002, as a consequence

of the Human Genome Project, PEREC-1 was discov-

ered to be TOMM40 — the outer mitochondrial mem-

brane channel through which peptides and proteins are

imported into mitochondria to support mitochondrial

function and biogenesis [28]. Because TOMM40 is

located immediately adjacent to APOE on the genome,

rather than simply assuming that the association between

TOMM40 and AD just reflected linkage to the corre-

sponding APOE alleles, both TOMM40 and APOE could

be viewed as individually important in a mechanistic

pathway to slowly damage mitochondrial dynamic



functions. This view of two genes that were in linkage

disequilibrium and encoded biologically interacting

proteins did not permeate the AD community even after

phylogenetic analyses demonstrated APOE4, APOE3,

and APOE2 alleles were linked to different alleles of a

polyT locus, rs10524523 (henceforth, ‘523’), in intron 6 of

TOMM40. In Caucasians these 523 alleles are defined as

Short (S), Long (L), and Very Long (VL). Genotypes of

the different polyT variant lengths at the 523 locus have

been demonstrated to have different age of onset distri-

butions. The APOE genotypes are fully represented

using TOMM40-523 genotypes (Figure 3).

A group of researchers led by Dr. Yadong Huang at the

Gladstone Institute [San Francisco, CA] investigated the

dynamic functions of mitochondria in neuronal tissue

cultures well before any AD GWAS data were published

[21��]. They also investigated the role that intraneuronal

apoE protein isoforms played in these dynamic functions,

Figure 3

L/L

L/VL L/S

S/S

S/VL

VL/VL

ε4/4

ε3/4

ε3/3

TOMM40’ 523 genotype curves colored by APOEgenotype:

Note that APOE3/4 and APOE3/3 are now informative

1.0

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0.070 80

Age

Fra

ctio

n w

ith

ou

t D

isea

se

90 100


Age at onset of cognitive impairment as a function of TOMM40-523

genotype. The Bryan Alzheimer’s Disease Research Center (ADRC)

Memory, Health, and Aging cohort (n = 508, 106 conversion events) was

followed prospectively at the Bryan ADRC at Duke University. Cognitive

status was determined using standard neuropsychological tests. The

age at which cognitive impairment occurred was retrospectively

stratified by TOMM40 genotype, and Kaplan–Meier curves were

constructed. TOMM40 genotype was determined using a sequencing

based genotyping method (Polymorphic DNA Technologies). TOMM40

genotypes and the corresponding APOE genotypes are indicated on the

figure. The red line corresponds to APOE e4/4; the two green lines

correspond to APOE e3/4, and the three blue lines correspond to APOE

e3/3. Note that within this cohort there were 78 individuals who carried

an APOE e2 allele, but only 5 developed cognitive impairment during the

study (VL/L (APOE e2/4), n = 1; S/L (APOE e2/4), n = 2; VL/VL (APOE

e2/3), n = 2). The number of subjects per 523 genotype (number

converted to case status) was: L/L, n = 23 (11); VL/L, n = 54 (24); S/L,

n = 72 (23); S/S, n = 100 (20); S/VL, n = 138 (22); VL/VL, n = 51

(6). Reprinted from [31]. The risk algorithm for qualification in the

TOMMORROW trial is illustrated in Ref. [31].


including the percentage of intraneuronal mitochondria

moving within a neuron, how far and how fast they moved

down the axon from the cell body to the terminal neurite

area, the differential interaction of apoE4, apoE3, and

apoE2 on these functions, and the effect of mitochondrial

acting drugs in correcting dynamic functions and increas-

ing neurite outgrowth [29��,30,31]. Earlier studies had

also demonstrated the interactions of APOE, APP, and

mitochondria [21��,22�,32].

In 2005, our research team at GlaxoSmithKline initiated

phylogenetic mapping experiments looking directly at

the LD regions on chromosome 19 that contains both

APOE and TOMM40. In 2007, we reported that a 10Kb

region that contains the TOMM40 gene produced a

robust phylogenetic tree structure [33] and we later

validated the phylogenetic tree structure in a different

cohort and further demonstrated that there was a specific

linkage between APOE and TOMM40 alleles (32). In

subsequent experiments, it became clear that the

TOMM40 region containing the intronic, variable polyT

marker, rs10524523, contributed independently of APOE

SNPs to the GWAS association results. Variations in the

polyT lengths differentiated terminal clades on the phy-

logenetic tree and specific polyT lengths were differen-

tially linked to downstream APOE genotypes. For

example the APOE3 allele can be linked to one of two

polyT alleles of rs10524523.

Phylogenetic mapping is a method that is commonly used

to detect new evolutionary mutations in the viral genetics

field. It has been used to map new mutation sites and the

evolution of new strains of influenza viruses so that new

vaccines can be manufactured annually. However, before

the consensus human genome data were available, the

methodology had not been applied to the analysis of

human genetic complex disease loci. It is not a GWAS

methodology, but rather a deep focus on a small region

that has suspected importance. Critically, if a region of the

genome shows low, to no, recombination, the approach is

potentially very powerful for genotype/phenotype associ-

ation analysis, based on the assumption that evolutiona-

rily closely related sequences will share phenotypically

important mutations.

In 2009, the results of two independent phylogenetic

mapping experiments of a small (10 Kb) region of the

genome located within the larger LD region on chromo-

some 19 that contains both the APOE and TOMM40

genes were published [34��] (Figure 4). In an IRB-

approved study of AD patients and age matched controls

from a cohort from the Arizona Alzheimer’s Disease

Research Center, DNA was sequenced and underwent

phylogenetic mapping. It was found that virtually all

patients with the APOE4/4 genotype segregated into a

specific branch, named Clade A. Clade B was almost

completely (98%) homogeneous for the APOE3 allele.


86 Neurosciences

Figure 4

Very long (30-36)linked to e3

(26-30) e4

(19-23) e4

Short (14-16)linked to e3

(a)

(b)

Rs10524523 poly-T polymorphisms evolved independently overtime and on differentgenetic backgrounds


0.0001

The sub-clades are differentiated by varying lengths of the poly-T allele (i.e. TOMM40-523). APOE3 can be linked to Very Long [Clade A, top red box] or

Short 523 poly-T alleles (Clade B, green box] APOE4 is connected to Long poly-T alleles (Clade A, 2nd and 3rd red boxes,] and a small group of shorter

Long polyT alleles [also Clade A, blue box]. Note especially that the APOE VL [Clade A] and the APOE3 Short [Clade B] clustered separately in distinct

phylogenetic branches.

APOE3/3 subjects segregated to both Clades A and B, as

did heterozygotes with the APOE3/4 and APOE2/3 gen-

otypes. The phylogenetic tree was further distinguished

by terminal clades, where clade membership was deter-

mined by the length of the polyT variant, rs10423523, of

TOMM40. The boxes illustrated in Figure 4 show where

polyT length variants from rs10524523 were distributed

in clades by different size ranges. It was found that the

Long [L] polyTs were generally on the same DNA strand

as APOE4, while either a Short [S] or Very Long [VL]

polyT strand was attached to APOE3 or the relatively

uncommon APOE2. Most of the haplotypes containing L

alleles of 523 were located in Clade A, but S or VL alleles

of TOMM40 were attached to APOE3 or APOE2 strands

and were divided to evolutionary-related branches be-

tween Clade A [VL] or Clade B [S]. Since S or VL variants

of 523 were located on the same strands as APOE3 or

APOE2, all other compound APOE genotypes containing

APOE3 or APOE2 alleles were distributed between the

major clades (Figure 3).

A prospective clinical onset study performed over two

decades, with defined age of onset recorded, was

examined retrospectively using the 523 genotypes.

Kaplan–Meier analysis of 523 and APOE genotypes again

showed that the APOE4/4 survival curve was identical to


523 L/L (Figure 3). The age of onset distribution curves

from a large prospectively collected clinical population

with accurate age of onset ascertainment demonstrated

that the APOE4/4 curve reflected the 523 L/L carriers

[35��] (Figure 3). The APOE3/4 distribution now could

be mapped as two age of onset distribution for 523 L-VL

and L-S, and the APOE3/3 and APOE2 carriers were

mapped to three curves, 523 S-S, S-VL or VL-VL. The

523 S or VL alleles attached to APOE2 are similar to

APOE3, but APOE2 carriers shift the age of onset to 8–10

years older. (9) In this way, the phylogenetic mapping

studies were pivotal in being able to elucidate the associ-

ation between TOMM40 rs10524523 variants and risk of

AD at a given age.

Several published studies attempted to simply look at AD

onset data that were not determined from prospective,

longitudinal studies with defined age of onset definitions,

and thus lack the ability to differentiate genetic popu-

lations[36,37]. Figure 3 shows TOMM40 genotype-

specific differences between the rapid decline in the

proportion unaffected by dementia within a decade cen-

tered around 72–75 years of age [35��].

Using TOMM40 genotype provides age-dependent risk

prediction for everyone as opposed to the use of APOE



where only for the 2% of the Caucasian population that

carry APOE 4/4 could an acceptable risk prediction be

made. A biomarker risk algorithm, based on TOMM40

genotype, APOE carriage, and age at presentation has

been developed that determines risk for clinical onset in

97% of the population. [Not enough subjects with the

APOE2/4 or 2/2 genotypes, about 3% of the population,

who developed AD were available from the prospective

Caucasian study to produce the necessary age distribution

curves] [35��]. The TOMM40 algorithm derived from the

age of onset distributions will be qualified for clinical

applications during the TOMMORROW study. Using a

validated testing methodology and prospective data, the

clinical validation of the biomarker risk algorithm is one

primary endpoint of the trial [38]. The trial tests a low

dose of pioglitazone, a drug that has the potential to

increase neuronal mitochondrial content [39��], to deter-

mine whether it delays the onset of Mild Cognitive

Impairment due to AD in cognitively normal subjects

at high risk of developing disease within the period of the

trial [35��,40].

APOE and TOMM40 are in LD and can be clinically

used as a prognostic device in Caucasians. There is great

variability in the spectrum of 523 polyT lengths in other

ethnicities, so studies of the relationship between

TOMM40-523 genotypes and age of MCI of the AD

type, or AD, onset are needed to generalize the risk

algorithm to non-Caucasian ethnicities [5�]. Further stu-

dies may demonstrate that each 523 genotype may be

generalized and provide similar risk information across

ethnicities, as is observed in other genetic diseases across

ethnicities [5�].

Disease onset prediction using these validated technol-

ogies can only be tested prospectively over time. It is

important to note that the Phase III, TOMMORROW

study, can also be the substrate for traditional efficacy

PGX experiments to identify other specific, relevant

genetic factors that contribute to the response or failure

of response to pioglitazone for the treatment of MCI due

to AD. Thus the efficacy PGX markers for treatment of

MCI due to AD, or AD, in addition to the 523 genotype,

may provide prognostic treatment information for already

affected patients.

A very short follow-up on the use of HLA-B5701 as an informative test for risk of majorallergic reactions: translation to medicalqualification and the PGX-related datatranslating to a marketing success afterpatent expiryQualification of HLA-B5701 was performed in a Phase III

trial designed to prospectively qualify the marker for

clinical practice. [1] What has not been stated in the

literature is the creation of a commercial bonanza based

on the discovery and qualification of B5701. As HIV


treatment regimes have progressed to use cocktails of

drugs in combination preparations, the choice of abacavir

to form the reverse-transcriptase function with preferred

safety has led to continued use in formulations, and huge

commercial earnings. The establishment of safety PGX

led to a resurgence of abacavir sales vis-a-vis its newer

competitors that still had no informative diagnostic for

other side effects and adverse events. The subsequent

use of abacavir in combination therapies has far exceeded

the original commercial expectations. This is still the

poster child for human PGX qualified by a prospective

Phase III trial. The approximate cost of the trial was $20

million, while the projected earnings are now extended

well past chemical patent expiry.

In summary, clinical pharmacologic imaging and PGX

uses are becoming intermixed in the vocabulary of drug

development and commercial potential. The safety PGX

that began in 1997, as a condition of accelerated market

approval, revived abacavir use. The benefits of both the

efficacy and safety of abacavir are still a source of com-

mercial value — never anticipated during develop-

ment — and are a victory for PGX by any measure.

The integration of new genetic technologies and PGX

methods into the drug development enterprise is

uncommon, except in cancer and infectious disease

where efficacy can be clinically measured quite accu-

rately and in shorter trials. Multiple genetic technologies

were used to move forward with the TOMMORROW

trial into uncharted drug development territory. Some

very expensive popular technologies, like GWAS, may

have been over-interpreted by the AD field. Hundreds

of new genes have been nominated for disease

associations by borderline statistical scores, yet to date

phylogenetic mapping using quantitative clinical

characteristics of disease has not been widely adopted,

despite translational success. The TOMMORROW trial

will qualify the genetic biomarker and test whether a

known, safe drug given at low dose can be effective in

delaying the onset of MCI due to AD in individuals at

high genetic risk for their specific ages. Prospective

delay of onset studies takes a long time to complete.

In the interim, treatment studies for MCI-AD of shorter

duration are being planned. Applications of phyloge-

netic analyses to other complex diseases of large medical

need, such as obesity and schizophrenia, are currently

underway.

References and recommended readingPapers of particular interest, published within the period of review,have been highlighted as:

� of special interest�� of outstanding interest

1. Hughes AR et al.: Pharmacogenetics of hypersensitivity toabacavir: from PGx hypothesis to confirmation to clinicalutility. Pharmacogenom J 2008, 8:365-374.


http://refhub.elsevier.com/S1471-4892(13)00234-8/sbref0005



88 Neurosciences

2. Roses AD: Pharmacogenetics in drug discovery anddevelopment: a translational perspective. Nat Rev Drug Discov2008, 7:807-817.

3. (CDER), C.f.D.E.a.R., C.f.B.E.a.R. (CBER), and C.f.D.a.R.H.(CDRH), Draft Guidance: Enrichment Strategies for Clinical Trialsto Support Approval of Human Drugs and Biological Products,U.S.D.o.H.a.H. Services and F.a.D. Administration, Editors. 2012,Office of Communication, Outreach and Development v http://www.fda.gov/BiologicsBloodVaccines/GuidanceComplianceRegulatoryInformation/default.htm.

4. (CDER), C.f.D.E.a.R. and C.f.B.E.a.R. (CBER), Guidance forIndustry: E15 Definitions for Genomic Biomarkers,Pharmacogenomics. Pharmacogenetics, Genomic Data andSample Coding Categories. U.S.D.o.H.a.H. Services and F.a.D.Administration, Editors. 2008: www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm073162.pdf.

5.�

Linnertz C et al.: Characterization of the poly-T variant in theTOMM40 gene in diverse populations. PLoS ONE 2012, 7:e4.

Illustrates high variability of the polyT allele fequencies in different ethnicgroups 3099.

6. McClellan J, King MC: Genetic heterogeneity in human disease.Cell 2010, 141:210-217.

7.��

McCarthy MI et al.: Genome-wide association studies forcomplex traits: consensus, uncertainty and challenges. NatRev Genet 2008, 9:356-369.

Questions raised about the conclusions of large GWAS studies and theirrelationship to biology.

8.��

Manolio TA et al.: Finding the missing heritability of complexdiseases. Nature 2009, 461:747-753.

Perhaps the missing heritability is not SNPs but more highly complexpolymorphisms.

9.�

Saunders AM et al.: Association of apolipoprotein E alleleepsilon 4 with late-onset familial and sporadic Alzheimer’sdisease. Neurology 1993, 43:1467-1472.

First report of APOE4 association to sporadic Alzhimer’s Disease.

10.��

Corder EH et al.: Gene dose of apolipoprotein E type 4 allele andthe risk of Alzheimer’s disease in late onset families. Science1993, 261:921-923.

First demonstration of relationship of APOE genotypes to age of AD onsetdistributions.

11. Peacock ML, Fink JK: ApoE epsilon 4 allelic association withAlzheimer’s disease: independent confirmation usingdenaturing gradient gel electrophoresis. Neurology 1994,44:339-341.

12. Ben-Shlomo Y, Lewis G, McKeigue PM: ApolipoproteinE-epsilon 4 allele and Alzheimer’s disease. Lancet 1993,342:1310.

13. Lucotte G et al.: Apolipoprotein E-epsilon 4 allele andAlzheimer’s disease. Lancet 1993, 342:1309.

14. Czech C et al.: Apolipoprotein E-epsilon 4 allele andAlzheimer’s disease. Lancet 1993, 342:1309.

15. Amouyel P et al.: Apolipoprotein E-epsilon 4 allele andAlzheimer’s disease. Lancet 1993, 342:1309.

16. Mayeux R et al.: The apolipoprotein epsilon 4 allele in patientswith Alzheimer’s disease. Ann Neurol 1993, 34:752-754.

17. Harold D et al.: Genome-wide association study identifiesvariants at CLU and PICALM associated with Alzheimer’sdisease. Nat Genet 2009, 41:1088-1093.

18. Hollingworth P et al.: Common variants at ABCA7, MS4A6A/MS4A4E, EPHA1, CD33 and CD2AP are associated withAlzheimer’s disease. Nat Genet 2011, 43:429-435.

19. Sleegers K et al.: The pursuit of susceptibility genes forAlzheimer’s disease: progress and prospects. Trends Genet2010, 26:84-93.

20. Gerrish A et al.: The role of variation at AbetaPP, PSEN1,PSEN2, and MAPT in late onset Alzheimer’s disease. JAlzheimers Dis 2012, 28:377-387.


21.��

Chang S et al.: Lipid- and receptor-binding regions ofapolipoprotein E4 fragments act in concert to causemitochondrial dysfunction and neurotoxicity. Proc Natl AcadSci U S A 2005, 102:18694-18699.

Effect of apolipoprotein E isoforms on mitochondrial functions: bindingand dynamic effects.

22.�

Devi L et al.: Accumulation of amyloid precursor protein in themitochondrial import channels of human Alzheimer’s diseasebrain is associated with mitochondrial dysfunction. J Neurosci2006, 26:9057-9068.

First demonstration that APP is accumulated with the mitochondrialimport channel, Tom40, supporting the interaction of Tom40 may berelated to effects attributed to amyloid cascade.

23.��

Reiman EM et al.: Preclinical evidence of Alzheimer’s disease inpersons homozygous for the epsilon 4 allele for apolipoproteinE. N Engl J Med 1996, 334:752-758.

First demonstration that APOE genotypes affect glucose utilization in thebrain decades before the onset of clinical symptoms of AD.

24.�

Small GW et al.: Early detection of Alzheimer’s disease bycombining apolipoprotein E and neuroimaging. Ann N Y AcadSci 1996, 802:70-78.

Rapid confirmation of APOE genotypes affecting glucose utilization.

25. Small GW et al.: Apolipoprotein E type 4 allele and cerebralglucose metabolism in relatives at risk for familial Alzheimerdisease. JAMA 1995, 273:942-947.

26.��

Roses AD et al.: Morphological, biochemical, andgenetic support for an apolipoprotein E effect onmicrotubular metabolism. Ann N Y Acad Sci 1996,777:146-157.

Early review of the relationship of apoE isoforms to effects on measure-able metabolic functions necessary for neuronal homeostasis.

27.�

Lai E et al.: A 4-Mb high-density single nucleotidepolymorphism-based map around human APOE. Genomics1998, 54:31-38.

First modelling of GWAS: Would GWAS have found APOE associationwith AD if GWAS had been able to be performed earlier?.

28. Strausberg RL et al.: Generation and initial analysis ofmore than 15,000 full-length human and mouse cDNAsequences. Proc Natl Acad Sci U S A 2002, 99:16899-16903.

29.��

Brodbeck J et al.: Rosiglitazone increases dendritic spinedensity and rescues spine loss caused by apolipoprotein E4 inprimary cortical neurons. Proc Natl Acad Sci U S A 2008,105:1343-1346.

Demonstration of the increased neurite outgrowth by rosiglitazonw,rescues the adverse effects of apoE on neurite outgrowth.

30. Mahley RW, Huang Y, Weisgraber KH: Detrimental effects ofapolipoprotein E4: potential therapeutic targets in Alzheimer’sdisease. Curr Alzheimer Res 2007, 4:537-540.

31. Chen HK et al.: Small molecule structure correctors abolishdetrimental effects of apolipoprotein E4 in cultured neurons.J Biol Chem 2012, 287:5253-5266.

32. Sanan DA et al.: Apolipoprotein E associates with beta amyloidpeptide of Alzheimer’s disease to form novel monofibrils.Isoform apoE4 associates more efficiently than apoE3. J ClinInvest 1994, 94:860-869.

33. Roses AD: ‘‘Other’’ AD-associated genetic variants – not justSNPs. Human Genome Variation and Complex Genome Analysis.2007:. Barcelona, Spain.

34.��

Roses AD et al.: A TOMM40 variable-length polymorphismpredicts the age of late-onset Alzheimer’s disease.Pharmacogenom J 2010, 10:375-384.

First application of human genome phylogenetics to discover a singlelocus with highly variable polyTs associated with compound genotypeassociations with age of onset of AD.

35.��

Crenshaw DG et al.: Using genetics to enable studies on theprevention of Alzheimer’s disease. Clin Pharmacol Ther 2013,93:177-185.

Phase III pharmacogenetic-assisted clinical trial to qualify a geneticalgorithm and test the effect of low dose pioglitazone to delay the onsetof Mild Cognitive Impairment due to AD.





http://www.fda.gov/BiologicsBloodVaccines/GuidanceComplianceRegulatoryInformation/default.htm



http://www.fda.gov/downloads/drugs/guidancecomplianceregulatoryinformation/guidances/ucm073162.pdf

































































































36. Cruchaga C et al.: Association and expression analyses withsingle-nucleotide polymorphisms in TOMM40 in Alzheimerdisease. Arch Neurol 2011, 68:1013-1019.

37. Jun G et al.: Comprehensive search for Alzheimer diseasesusceptibility loci in the APOE region. Arch Neurol 2012,69:1270-1279.

38. Lutz MW et al.: Performance of a biomarker risk algorithm for aclinical trial to delay onset of mild cognitive impairment due to


Alzheimer’s disease. 2012 Annual Meeting of the AmericanSociety of Human Genetics; San Francisco, CA: 2012.

39.��

Strum JC et al.: Rosiglitazone induces mitochondrialbiogenesis in mouse brain. J Alzheimers Dis 2007, 11:45-51.

Increasing mitochondrial biogenesis as a mechanism of action toincrease neuritic outgrowth.

40. Roses AD: An inherited variable poly-T repeat genotype inTOMM40 in Alzheimer disease. Arch Neurol 2010, 67:536-541.
















new drug development d roses , 2, ann m saunders , michael ...drsmith/migrated/metx270/html/roses et...

Documents